332.2M - 333.9M (aka 100M digit range)

[LaurV]

I saw some activity in 334M++, someone is taking all from 65 to 66 bits. I just want to make sure he does not work in the lower range below 334M. I am taking all 332M-334M range from 66-7-8 bits to 69 bits, and I want to make sure I don't step on someone's shoes.

Yesterday I re-arranged my workers (remote, I am still in holiday till 16 and don't have full access to the computer). Before, I was splitting the 332M-334M range in 6 slices, as it was the easiest "cut" to do in the over-30k lines of the worktodo file. My first 3 worker's activity can be "guessed" in the first part of the range (other people working there too, so my activity is not immediately visible, but you can deduct it if you count the expos and factors), but is can be seen exactly, especially in the second part of the range, where no one else is working, so it is quite clear where each of my last 3 workers starts its range.

As I said, yesterday I re-arranged the worktodo files, because some reports came back with "result for this expo was not needed" (about 8 times from few hundred expos). So, some random guy reserved some expos there and did the work. That is why I re-downloaded the list of unreserved exponents factored lower then 69 in the 332-334M range. But this time I made the split in a cleverer way: I did a small batch to read the file, put line1 into w1.txt, line2 into w2.txt, line3 into w3.txt... line6 into w6.txt, then again, line7 into w1.txt, line8 into w2.txt... etc. I deleted the old worktodo files and replaced them with w1, w2,... w6. Now the workers will work the lowest possible exponent. They already finished the "under-100M-digits" range (of 332.1M, about 1000 expos to 69 bits) yesterday in few hours and moved to 332.5 (as per Uncwilly's suggestion), and finished those overnight, too. The reports are not yet updated on mersenne.info, but there should be no exponent between 332M0 and 332M6 factored under 69 bits (except maybe if some of them were reserved and they are not part of my task).

[c10ck3r]

Originally Posted by LaurV

I saw some activity in 334M++, someone is taking all from 65 to 66 bits. I just want to make sure he does not work in the lower range below 334M. I am taking all 332M-334M range from 66-7-8 bits to 69 bits, and I want to make sure I don't step on someone's shoes.
/SNIP/

I didn't think I would be that noticeable :)
I'm running 2 instances of mfaktc 0.17 on a GTX460 in the 334,xxx,xxx range. Not going higher or lower until the other 150kb of worktodo.txt are done, then will consider new ranges.

[LaurV]

332M5 and 332M6 done to 69.
332M7 half through it to 69 according to last check. Is faster then I imagined, if nobody touched the computer during our absence. Unfortunately we just arrived back home.

[Uncwilly]

Originally Posted by Uncwilly

I see 47 exponents cleared in the last few days.
Also, now the 332M range has passed 57.6% of all exponents factored.

And now we are over 58.1% out on the range and have dropped ~300 exponents since March 25. Over 10,000 exponents have been taken from 67 and 68 to 69 or higher in the same period.

[LaurV]

I finished all 332M to 69 (appeared in today's report) and found 345 factors, preserving them in a file if someone need them gimme a sign.

I also attacked 333M and finished 333M0, starting 333M1 but I have to stop, from heat issues. I found 79 factors, put them with salt and brine in the same file as above, which contains 210 more factors from 333M2 to 333M7, with this range I played before holiday.

Well, as I said before, April is the hottest period of the year in this part of the world, there are over 40C outside and people spray water on each-other like crazy. Here inside the box the thermometer shows over 80C from time to time, when there are over 35 in the room and no wind around... and I am not going to let the aircond on when I go to the job. That is why I think I will take a break (read: reduce the load, at least) till the beginning of May, when the rains will start and the temperature will go down.

[Uncwilly]

Originally Posted by LaurV

That is why I think I will take a break (read: reduce the load, at least) till the beginning of May, when the rains will start and the temperature will go down.

That's cool.
Thanks for all of your efforts in this range. I have at times moderated my work because of cooling issues.

I gather the factors of exponents between 332192831 and 332599999. The bit depth, k, and effort equivalent get calculated for them. So far there are the following cases that are clearly P-1 finds: 84.8 bit, 86.9 bit, 87.8 bit, 88.1 bit, 91.3 bit, 92.4 bit, & 102.3 bit. These would represent the following GHz-days, if done by T-F, respectively: 184, 11,792, 47,166, 94,333, 188,665, 1,509,320, 3,018,641, 3,091,087,963.

[LaurV]

None of those astronomical numbers are mine, I strictly did TF from 66-7-8 to 69 bits, and very few expos to 70 bits. I don't know who did those.

[Uncwilly]

Here is the 'monthly' status report for the range from 332192831 to 332399999:

Date of data				4/20/2012
Average bit depth for first 100 expos	78.40
Average bit depth for first 1000 expos	77.16
100th active expo (no factor found)	332198357
1000th active expo (no factor found)	332246137
Unitless total effort number		173,957,120
Number of first 100 expos to 2^77	88
Number of first 1000 expos to 2^76	883
Number left in range			4217

Estimated expos in range to be removed	283
(by taking all expos to 2^79)

Bit	# at bit level
71	988
72	1160
73	339
74	302
75	133
76	545
77	286
78	30
79	405
80	24
81	4
P-1     182

2 of the numbers that had factors found were a 79 and an 80 bit. This hurt the effort value by 786,432. Attached is a graph that shows progress over time and what portion comes from the various bit levels.

[LaurV]

I finished all the range of 332M-334M to 69 bits.

During this work, which took about a month on an average of 2 GTX580 (sometime 3, sometime 1, sometime none :D), I found about 1400 factors. If anyone need them for whatever reason, they are attached here below.

• 345 factors for the first half of the range (332M to 333M). In this range the most of the exponents were already to 68 bits, the rest were to 67 or 66, and other people also did a lot of work before. They were about 20 or 30 factors more in the beginning, until I become enough clever to start recording them... (they were reported to the server, but I did not record them in the file)

• 1043 factors for the second part of the range (333M to 334M). Here all exponents were to 66, except maybe one or two which were to 67, not so many work done before, that is why more factors were found.

Note that the factors may not be singular (I stopped when a factor was found, with few exceptions when the factor was found in the LAST class, they do not have the "star") and may not be the smallest in the range of 66-69 bits (for ppl who don't knows how mfaktc works: it sieves by modularity classes, so it can find a bigger factor before finding a smaller one, if the class of the bigger factor is smaller then the class of the smaller one). Also some other known factors may exists, these are NOT "all factors", but only the factors I found. They can be useful for whatever tests one may want to do.

[Uncwilly]

Originally Posted by LaurV

I finished all the range of 332M-334M to 69 bits.

Thanks for all of that help. Anytime that you have a free instance of mfactx, your help would be welcome.

[Uncwilly]

Here is the 'monthly' status report for the range from 332192831 to 332399999:

Date of data				5/21/2012
Average bit depth for first 100 expos	78.41
Average bit depth for first 1000 expos	77.16
100th active expo (no factor found)	332198357
1000th active expo (no factor found)	332246143
Unitless total effort number		174,375,936
Number of first 100 expos to 2^77	89
Number of first 1000 expos to 2^76	883
Number left in range			4214

Estimated expos in range to be removed	281
(by taking all expos to 2^79)

Bit	# at bit level
71	901
72	1214
73	344
74	328
75	133
76	538
77	292
78	30
79	405
80	24
81	5
P-1     188

The P-1's are keeping a fairly steady march.